如何使用分类词词典从文本片段中提取特定词？

Question

我想从数据帧中的文本中提取特定的单词。这些单词我已经输入到字典的列表中，它们属于某些类别（键）。由此，我想创建与存储单词的类别相对应的列。和往常一样，最好用例子来说明：

我有一个数据框：

df = pd.DataFrame({'Text': ["This car is fast, agile and large and wide", "This wagon is slow, sluggish, small and compact with alloy wheels"]}  )  

它创建表：

    Text
0   This car is fast, agile and large and wide
1   This wagon is slow, sluggish, small and compact with alloy wheels

还有一本我想从中提取的分类词词典。这些单词都是没有符号的自然语言单词，可以包括短语，例如本例中的“alloy wheels”（这不一定是字典，我只是觉得这是最好的方法）：

myDict = {
  "vehicle": ["car", "wagon"],
  "speed": ["fast", "agile", "slow", "sluggish"],
  "size": ["large", "small", "wide", "compact"]
  "feature": ["alloy wheels"]
}

根据这个，我想创建一个如下所示的表：

|     Text                                                          | vehicle | speed          | size           | feature      |
| ----------------------------------------------------------------- | ------- | -------------- | -------------- | ------------ |
| This car is fast, agile and large and wide                        | car     | fast, agile    | large, wide    | NaN          |
| This wagon is slow, sluggish, small and compact with allow wheels | wagon   | slow, sluggish | small, compact | alloy wheels |

为提前求助干杯！很想使用正则表达式，但欢迎任何解决方案！

Answer 1

有很多方法可以解决这个问题。我可以从一个方法开始：定义一个函数，如果它们与你的句子匹配，它将返回一个单词列表。

    def get_matching_words(sentence, category_dict, category):
        
        matching_words = list()

        for word in category_dict[category]:
             if word in sentence:
                   matching_words.append(word)

        return matching_words

然后，您希望将此函数应用于数据帧。

    df["vehicle"] = df.apply(lambda x: get_matching_words(x, "vehicle", my_dict)

    df["speed"] = df.apply(lambda x: get_matching_words(x, "speed", my_dict)

这里要添加的唯一内容是将列表连接成字符串，而不是返回列表。

def get_matching_words(sentence, category_dict, category):
        
        matching_words = list()

        for word in category_dict[category]:
             if word in sentence:
                   matching_words.append(word)

        return ",".join(matching_words)

希望这有帮助